Monitoring Agentic Systems Before They're Reliable
2026-06-01 • Software Engineering
Software EngineeringArtificial Intelligence
AI summaryⓘ
The authors studied systems made of multiple parts that often fail due to structural problems rather than mistakes in specific tasks. They created a method to watch these systems from three angles—quality, suitability, and efficiency—across different monitoring levels to spot problems better. Their tests showed that structural issues hide task-level errors, making those errors hard to detect early on. They also developed a way to prioritize which problems need human attention and found that most can be tracked automatically. They suggest that as systems get better integrated, monitoring should evolve from spotting structural flaws to tracking specific task errors and overall reliability.
agentic systemsstructural defectstask-level errorsmonitoring scopesvariance characterizationFMEA (Failure Modes and Effects Analysis)triage methodologycross-run monitoringintegration gapsmulti-stage workflows
Authors
Marisa Ferrara Boston, Glen Hanson, Effi Georgala, JD Hudgens, Heather Frase
Abstract
Agentic systems entering production typically operate as partially integrated assemblies where structural defects, not task-level errors, dominate the failure landscape. At this maturity level, task-level error detection may be infeasible: structural failure modes mask the signal that task-level monitors are designed to detect.We present a monitoring and triage methodology that decomposes agentic system evaluation into three dimensions (quality, suitability, efficiency) at three monitoring scopes (within-run, cross-run, structural), using variance as a characterization signal. Findings are routed through severity classification adapted from FMEA, concentrating human attention on the subset that warrants investigation. We evaluate on a synthetic testbed of 220 runs across 120 document bundles with controlled error injection.Three results emerge. Monitor scope determines failure type: within-run monitors surface deterministic stage defects (CV = 0.02), cross-run monitors surface stochastic integration consequences (CV = 1.25, 24% at L2), and a structural monitor identifies an integration gap with perfect consistency (CV = 0.00). Injected task-level errors are indistinguishable from clean baselines, confirming structural defects mask task-level signal. Deterministic triage routes 97% of findings to automated tracking, leaving the 2% reflecting variable behavior for human investigation.We propose, on Stage 1 evidence, a maturity-staging model in which monitoring transitions from structural characterization to error detection to reliability tracking as integration defects resolve. The taxonomy, CV-based scope characterization, and severity model transfer architecturally to document-driven, multi-stage agentic workflows in regulated industries; specific calibrations are domain-specific. Deploy monitoring early: the first thing it finds is the most important thing to fix.